Provided are method and apparatuses related to material recognition and training. A training apparatus for material recognition generates training data associated with a material by generating a texture image having a texture attribute from an object image and recognizing material information from the texture image using a material model.
Legal claims defining the scope of protection, as filed with the USPTO.
1. A processor-implemented material recognition method of a computing apparatus, the method comprising: acquiring an input image comprising an object; excluding shape information of a shape associated with the object based on the input image; generating a texture image indicating a texture of the object based on the input image excluding the shape information of the shape; recognizing material information from the generated texture image using a trained material recognition model; and controlling the computing apparatus based on a result of the recognizing.
2. The material recognition method of claim 1 , wherein the generating comprises: extracting a texture region from the input image; and generating the texture image by extending the texture region to a predetermined size.
3. The material recognition method of claim 2 , wherein the extracting comprises: determining an object region from the input image; and extracting at least a portion of the object region as the texture region.
4. The material recognition method of claim 3 , wherein the extracting of at least a portion of the object region comprises extracting a region including a central point of the object region as the texture region.
5. The material recognition method of claim 3 , wherein the extracting of at least a portion of the object region comprises extracting a region from the object region at random as the texture region.
6. The material recognition method of claim 2 , wherein the extracting of the texture region comprises: segmenting the input image into patches; calculating a similarity between the patches; and extracting a region having patches with a similarity greater than or equal to a threshold as the texture region.
7. The material recognition method of claim 6 , wherein the calculating of the similarity between the patches comprise generating intensity histograms for the patches, and calculating a similarity between the intensity histograms of the patches.
8. The material recognition method of claim 1 , wherein the generating comprises extending a texture region extracted from the input image.
9. The material recognition method of claim 1 , wherein the generating comprises extending a texture region extracted from the input image to maintain a resolution of the texture region to be greater than or equal to a resolution of the input image.
10. The material recognition method of claim 1 , wherein the generating comprises extending a texture region extracted from the input image to a size greater than or equal to a size of the input image.
11. The material recognition method of claim 1 , wherein the generating comprises extending a texture region extracted from the input image to maintain a texture attribute to be greater than or equal to a threshold.
12. A non-transitory computer-readable medium storing instructions that, when executed by a processor, causes the processor to perform the method of claim 1 .
13. A processor-implemented training method for material recognition, the method comprising: extracting a texture region from an object image; generating a texture image by extending the texture region; and generating training data by excluding shape information of object shapes from the training data and by mapping material information to the texture image; and training a material recognition model using the training data for the material recognition.
14. The method of claim 13 , further comprising: training the material recognition model to output the material information from one of the texture image or the object image based on the training data.
15. The method of claim 13 , wherein the extracting comprises: determining an object region from the object image; and extracting at least a portion of the object region as the texture region.
16. The method of claim 15 , wherein the extracting of at least a portion of the object region comprises extracting a region including a central point of the object region as the texture region.
17. The method of claim 15 , wherein the extracting of at least a portion of the object region comprises extracting a region from the object region at random as the texture region.
18. The method of claim 13 , wherein the extracting of the texture region comprises: segmenting the object image into patches; calculating a similarity between the patches; and extracting a region having patches with a similarity greater than or equal to a threshold as the texture region.
19. The method of claim 13 , wherein the generating comprises extending the texture region to a predetermined size.
20. The method of claim 13 , wherein the generating comprises extending the texture region to maintain a resolution of the texture region to be greater than or equal to a resolution of the object image.
21. The method of claim 13 , wherein the generating comprises extending the texture region to a size greater than or equal to a size of the object image.
22. The method of claim 13 , wherein the generating comprises extending the texture region to exclude a visual defect.
23. A computing apparatus, comprising: a memory configured to store a trained material recognition model; and a processor configured to: acquire an input image comprising an object; exclude shape information of a shape associated with the object based on the input image; generate a texture image indicating a texture of the object based on the input image excluding the shape information of the shape; recognize material information from the generated texture image using the trained material recognition model; and control the computing apparatus based on a result of the recognizing.
24. A training apparatus for material recognition, the apparatus comprising: a processor configured to: extract a texture region from an object image; generate a texture image by extending the extracted texture region; generate training data by excluding shape information of object shapes from the training data and by mapping material information to the texture image; and perform a training of a material recognition model using the training data for the material recognition; and a memory configured to store the training data and the trained material recognition model.
25. A digital device comprising: a sensor configured to capture an image of an object; a memory; and a processor configured to: mask a shape attribute of the object, by extracting a texture region from the image of the object; generate a texture image by extending the extracted texture region; generate training data by excluding shape information of object shapes from the training data and by mapping material information to the texture image; train a neural network using the training data to output the material information, in response to the texture image being input; and store the training data and the trained neural network for material recognition in the memory.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
May 11, 2017
March 31, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.